现阶段人工智能应用于我国基层眼病筛查中的新挑战

Emerging challenges in the application of artificial intelligence for the eye disease screening in Chinese primary healthcare institutions

人工智能(AI)应用于我国基层眼病筛查近期取得较大突破, 但也显现出新挑战。一是AI软件不断突破, 筛查眼病种类不断扩大, 筛查诊断准确性不断提升, 并逐渐向预测眼病进展发展。但5G等基础设施不足和专业人员短缺影响了筛查的覆盖率。二是AI经济性已明确, 但新的筛查模式也影响到了筛查的公平性。需要关注AI应用模式"中国化"。三是AI筛查相关指南不断健全, 指明了AI发展方向, 为AI技术的推广和应用提供了参考依据。但亟待高质量实证研究为AI应用于基层眼病筛查的政策制定提供科学证据。因此, 笔者建议发展结合症状和病史等基础资料以及简易眼科检查的多模态AI模型;加速5G等基础设施建设, 注重医工交叉人才培养;因地制宜探索大规模眼病筛查的适宜服务体系和模式;开展长时间、大规模、多中心的实证研究。

Breakthroughs have been achieved recently in the application of artificial intelligence(AI)for the eye disease screening in Chinese primary healthcare institutions,but challenges have also emerged.First,AI software has continuously evolved,expanding the range of eye diseases that can be screened,enhancing diagnostic accuracy,and progressing towards predicting the course of eye diseases.However,inadequate infrastructure such as 5G and a shortage of specialized personnel have hindered the coverage of screenings.Second,while the cost-effectiveness of AI is well-established,new screening models have impacted the equity of screenings.It is essential to tailor AI application models to the specific context of China.Third,AI screening guidelines have been increasingly improved,providing direction for AI development and reference for the promotion and application of AI technologies.Nonetheless,high-quality empirical research is urgently needed to provide scientific evidence for policymaking related to AI in the eye disease screening.Therefore,it is suggested to develop multimodal AI models that integrate basic data such as symptoms and medical history with simple ophthalmic examinations,to accelerate the construction of infrastructure like 5G and focus on cultivating interdisciplinary talents,to explore suitable service systems and models for the large-scale eye disease screening tailored to local conditions,and to conduct long-term,multi-center,empirical studies.